We’re back! Harry, Rory and Jason!
The venture capital playbook is broken. Not bent — broken. In the latest 20VC x SaaStr episode, Harry Stebbings, Jason Lemkin, and Rory O’Driscoll dissect why even Sequoia is making dramatic leadership changes, why seed investing at $50M pre-money might not work anymore, and what it actually takes to build venture returns in the age of AI.
This isn’t your typical venture conversation about “exciting trends.” This is three investors with $3+ billion in combined AUM telling you what’s actually working, what’s spectacularly failing, and why the old playbook from 2015-2022 is now a liability.
Key Takeaways
On Venture Capital Evolution:
- Sequoia’s leadership transition reflects broader industry truth: most VCs and executives from the last decade aren’t the right people for the next decade
- The pace of AI evolution means knowledge from 6 months ago is probably wrong; staying current requires dedicated time investment
- Partnerships are inherently dysfunctional when performance can’t tie to economics, creating inevitable internal tension
On AI Investment Strategy:
- Only three ways to win: (1) Attach to compute budgets, (2) Replace human headcount, or (3) Massively displace incumbents
- Using AI to “make your product better” no longer earns any kudos — that’s table stakes in 2024
- Co-pilots were the 2024 story that didn’t work; agents becoming actual team members is the 2026 opportunity
On Deal Dynamics:
- Getting into deals at 5-10M ARR requires top-decile metrics — there’s almost no middle class of fundable companies
- The quality and speed of competitive clones has increased dramatically, compressing the window for building moats
- Traditional seed defensibility is dead; founders must run faster and bet on scale creating the moat, not early product advantages
On Portfolio Construction:
- With increased variance in AI deals, diversification becomes more critical, not less
- Small fund sizes ($40-100M) with acceptance of dilution can generate superior returns (10x+) versus large funds maintaining ownership (5x)
- 80+ company meetings per week per partnership is one approach; building deep relationships with fewer founders is another
On Fundraising Process:
- The best fundraises don’t feel like processes — they’re cultivated over months with 3-4 investors ready before the data room opens
- Taking a term sheet immediately versus “running a process” depends on capital efficiency and relationship quality
- Founders often overlearn “run a process” advice without understanding the optimal approach is having everyone ready to commit before you formally raise
On Market Dynamics:
- Companies attached to AI compute infrastructure (like DataDog) are crushing it; those just using AI for product improvement (like Duolingo) are getting punished
- The $100M ARR milestone with 50 people (like Gamma) represents a new efficiency paradigm
- Capital-efficient outcomes (Billion to One at $5B, Navan at $4.5B) deliver superior investor returns despite smaller headline valuations
Sequoia’s Move: What It Really Means
When Sequoia replaced Roelof Botha as managing partner after just three years with Pat Grady and Alfred Lin in a split leadership structure, the venture world noticed. But the reaction from our panel was telling: this wasn’t about interpersonal drama or normal succession planning.
“Whenever you have a CEO change happening in venture, it’s because something isn’t working,” Rory stated bluntly. “This is dissatisfaction about how the firm is doing relative to the competition. They missed some rounds and some deals. They passed on some great companies.”
But rather than viewing this as a Sequoia-specific challenge, the panel sees it as symptomatic of a broader industry issue.
“More people should be stepping aside today,” Jason argues. “VCs, executives, founders. Most folks from the last decade or 15 years are not the right people for the next decade. I genuinely believe it across my ecosystem. I struggle to even recommend a lot of the CROs and executives I know for roles today.”
Rory adds context: “The pace of evolution is so fast. If you decide what I knew 6 months ago is still useful, you’re probably going to be wrong very quickly. That’s what I find the most stressful about right now.”
The meta-lesson? Even the best firms are struggling to keep pace with AI, which means everyone else should be asking themselves hard questions about whether they’re still the right people for this moment.
And Sequoia gets credit for one thing: ruthless decisiveness. “They did not do that fatal error of saying it’s someone’s turn so we’ll leave him in,” Rory notes. “They ruthlessly said, if we’re going to compete, we need these people, not those people.”
Michael Burry’s $1.1B Bet Against Nvidia: Why Shorting AI is Brutally Hard
Michael Burry made headlines (again) by shorting Nvidia and Palantir to the tune of $1.1 billion. Unlike most pundits who just discussed whether he’d be right, Jason actually did the math on what it takes to make money on these bets.
“It brings home how hard a business it is to bet against AI capex,” Jason concludes. “You not just got to be right, but you got to be right on timing.”
Rory’s take: “It’s easy to be roughly right. It’s very hard to imagine that the AI capex boom doesn’t have a significant correction. But going from that kind of armchair podcast statement to actually being able to make money on it — that’s damn hard.”
The counterpoint to all this short-selling cynicism? The revenue is actually showing up. OpenAI is projected to hit $20B ARR this year, while Anthropic projects $70B ARR by 2028. The growth rates aren’t just holding — they’re accelerating.
Harry drives this home: “Are we being overly British? Are we looking for a problem that’s not there? The revenue is showing up in billions.”
The lesson: It’s intellectually easy to be a skeptic about AI valuations. It’s financially hard to make money being that skeptic. And meanwhile, the actual revenue keeps proving the bulls right.
Gamma’s $100M Revenue Run Rate with 50 People: The New Efficiency Paradigm
Few stories capture the AI revenue opportunity better than Gamma’s latest round. The company raised $100M at a $2.1B valuation, having hit $100M in revenue with just 50 people and 2 million users.
Jason and Team SaaStr are super-users and he walks through exactly how they use it at SaaStr: “Instead of sending sponsors the same dated prospectus, Gamma automatically pulls all of our data from Salesforce and our marketing automation system. It knows the exact number of leads and ROI. It knows who their competitors are and makes a fully dynamic piece of collateral in about 10 minutes.”
Here’s the math that’s important: SaaStr pays $100/month for Gamma — $1,200 per year. That’s stealth TAM expansion. “How much do we spend for Google Slides? Zero. It’s built in. How much do we spend on PowerPoint? I don’t even know where my key is to Microsoft Office. So it is a stealth TAM expansion.”
But Jason’s most important insight is about what’s coming next: “The lame thing about co-pilots is they were just tools. When the AI is part of your team for real, not VC talk, the amount of revenue that it’s accessible is so high.”
He distinguishes between AI as a tool versus AI as a team member: “It is sufficiently autonomous, knowledgeable and powerful to complete material high-value tasks on its own with some daily discussions just like on our team. The level of autonomy and capability — Gamma, go into my Google Calendar, create prospectuses and sales collateral for all 20 sales calls this week, pull all the data from Salesforce and HubSpot and Marketo, review once, and distribute to the team.”
The implication: We’re transitioning from AI making us more efficient (2024) to AI being actual members of the team (2026). That’s where the revenue explosion happens.
But there’s a warning embedded in Gamma’s story too. Canva now has a Gamma clone that’s “pretty good.” The competitive response time has collapsed from years to months or even weeks. This leads to one of the episode’s most important debates…
Does Defensibility Exist Anymore? The Most Important Question in Venture
This might be the single most important strategic question facing every seed and Series A investor right now. And the panel doesn’t fully agree on the answer.
Jason frames the problem: “The quality of clones is only going up. I can think of one investment I’ve made that has had five clones in the first 30 days, including one from a cloud leader. The ability of AI to enable us to clone better stuff faster — Canva is borderline competitive with Gamma and wasn’t when Cliff was on the show. What does seed investing mean when you might see 10 better versions in 30 days?”
Rory’s response: “I don’t think you can have a major defensibility at the seed or even frankly the stage we’re investing at. The defensibility theorem emerges at scale. Once you become the anointed winner, once a market coalesces and there’s two or three people, you know, it’s yours to lose.”
But Rory pushes back on the idea that this makes seed investing impossible: “You just have to internalize the game you’re in. Awesome team, run fast, be superlative on technology, get your distribution early, and then rely on that. You can’t be anointed the winner up front. Get over it everybody.”
Jason’s concern is economic: “Is that okay at 50 post for a seed round? Do the outcomes justify it? If I’ve got to spread $5 million checks around at 50 post, it’s tougher.”
Harry asks the critical question: “Are we getting paid for the risk?” He notes that even at the Series B stage, it’s unclear who will win in many categories. Looking at customer support AI companies that have raised Series B rounds: “I got no idea who the winner is in that category and I don’t think anyone does to be honest.”
The debate crystallizes around whether you can know enough at the B to justify the valuation:
Rory’s view: By the Series B, you can see rank order. “Once the horses are running and once they round the first furlong, you can actually see the rank order of where they’re running in a way you can’t at the early A.”
Harry’s counter: Looking at codegen — even with clear leaders like Cursor emerging, you have Codex making incredible ground, Claude Code advancing, and Replit and Lovable coming from different angles. “I think that’s still entirely up for grabs.”
The practical implication: This uncertainty means one of two things:
- Accept that seed/Series A investing requires higher diversification (40+ companies instead of 20)
- Or accept lower ownership percentages betting on outcome expansion
Jason’s conclusion: “Maybe you need 40 deals at $5M. That’s $200M for first checks. $200M for reserves. That’s $400M. $100M for fees and backup. I need at least $500M for my little seed fund to make the math work.”
The Three Ways to Actually Capture AI Revenue
Amid all the complexity, the panel crystalizes a simple framework for what actually works in AI investing right now.
Path 1: Attach to Compute Infrastructure
DataDog crushed earnings, with stock up 23% and $15M+ in AI-native customers. Why? They’re selling to the people making AI, and as those companies grow, DataDog sells more too.
Jason: “The AI leaders, the hyperscalers — they’re starting to buy like classic B2B companies. They’re recycling the same people in procurement. So if you’re attached to AI budget and you’re a DataDog era, you’re actually going to have a great 2026.”
Rory adds: “DataDog is a core piece of compute infrastructure and these hyperscalers are the most compute-intensive companies that have ever been known. If you’re selling compute stuff, you should be having a great quarter. If you’re selling routers, switches, interconnects, whatever it takes to stand up Stargate — you’re going to be golden.”
Path 2: Replace Human Headcount
“Where are you replacing humans for real?” Jason asks. “Where are you going to go in and reduce the headcount that vendor needs by half?”
This is the Replit story — Jason deployed 20+ AI agents that generate significant revenue while requiring substantial daily management but far less human headcount than traditional approaches.
The key distinction: “Most of the time, ‘replacing humans’ is a story. But when it’s real, when the AI agent is sufficiently autonomous to complete material high-value tasks, that’s when you’re actually part of the team, not just a tool.”
Path 3: Massively Displace an Incumbent
“You have a third option,” Jason concedes. “Use AI to massively displace an incumbent and steal all the revenue. In fact, that’s the history of B2B software mostly. I just don’t know how many of our public leaders are in a place to steal their own revenue.”
But he’s skeptical this is as compelling as the first two: “There’s like 400 AI CRM startups out there all saying they’re going to eat HubSpot’s and Salesforce’s lunch. That’s not exciting to me as an investment. That’s much less exciting than truly replacing 90% of your GTM team.”
If you’re not doing one of these three things, you’re probably not going to see explosive growth. Using AI to make your product better? That’s table stakes now, not a competitive advantage.
How to Actually Run a Fundraising Process (Without Running a Process)
One of the most tactically useful segments came when Harry shared a founder interaction that frustrated him: the company said they wanted to “run a process” despite Harry offering terms they’d stated they wanted.
Harry’s reaction: “I said, listen, if you’re running a process, you’re either optimizing for price or partner selection. I’m giving you a great price today that you said you wanted, which means you’re not optimizing for price. Which is just saying you think you can get better than me. In which case, just tell me straight.”
Rory sees both sides: “From the founder’s side, they’re correct. I see more failed financings because they didn’t run a process than because they did.”
But then he makes a critical distinction: “What typically happens when people ‘don’t run a process’ is someone comes in, says they’re interested, and the founder shares information serially to someone not yet ready to commit. That’s a mistake. That’s running an accidental process.”
Jason has the most nuanced take on optimal strategy: “The best founders cultivate enough interest with enough good VCs that if they hit the number, they just send an email. Harry, I’m thinking about raising around before the end of the year. And Harry could say, I’ll give you a term sheet today. And the right answer is: ‘I love you, Harry. I’m not ready today. I will be ready at the end of the year.'”
The key insight: “The optimal way to run it is for everyone already to want to invest for real without games before you open your data room. The best run processes don’t require a data room. Not a traditional one. Only one for diligence.”
Rory synthesizes: “The best run processes don’t feel like a process, but they are. If a founder is smartly nurturing relationships, keeping people broadly informed, then tries to time the interest such that when they’re ready to put their hand up, there’s three people who are primed and ready to go — that is perfection itself.”
But here’s the brutal caveat: Jason points out that this only works if you have stellar numbers. “You’re either YC, Neo, South Park Commons, got something, get funded — or who the hell is going to find you? You better be hot AI-native with top-quartile venture growth or you ain’t getting funded.”
Harry shares a data point that proves this: A company that grew from $400K to $3M ARR (classic enterprise SaaS, 10x growth) had 120 meetings and got one term sheet at $10M on $40M post. That’s 12x revenue for a 10x grower — which five years ago would have generated five term sheets from top firms.
“It’s the most binary fundraising environment in our lifetimes,” Jason concludes. “You’re either Captain Obvious getting funded, or you’re not seeing it anywhere.”
Duolingo Down 25%: The Cautionary Tale of the “Wrong” Kind of AI
While DataDog surged 23%, Duolingo crashed 25% in the same week. The contrast illustrates everything about what’s working and what’s not in AI.
Jason’s diagnosis: “Duolingo has the ‘wrong’ kind of AI. Duolingo is using AI to make its product better. Hooray. Every single portfolio company at scale should be using AI by this point to make your product better. This is not 2023. You don’t get any kudos for sprinkling AI dust on your product.”
Rory pushes back slightly on behalf of consumer apps: “If you’re an infrastructure company, you can co-attach to compute. If you’re a new AI apps company, you are using that compute. But if you’re like Duolingo — preexisting — what you can do is co-adopt the technology.”
But Jason sees a bigger problem: “Duolingo took money from Berlitz and all these language schools. Hooray, you did that. Now, where are you going to disrupt humans? This is your job. You already disrupted those humans. Unfortunately, they’re gone. Where’s the next level of human disruption?”
Rory makes an interesting point about where the AI opportunity in education actually is: LLM learning is equivalent to one-on-one human tutoring. What AI should enable in education is allowing everyone to get one-on-one learning instead of group-based learning. But there’s no budget in K-12 to give every kid a customized tutor. The market is adults wanting to learn a second language for business who would pay for a one-on-one coach — a niche market.
Jason’s framework applies here too: “If you’re not removing humans from the equation, you’re going to be heavily discounted. You’re either getting money from compute or you’re getting money because you’re using AI to replace humans. Otherwise, congratulations on your 14% growth.”
Portfolio Construction: The Math Actually Matters
One exchange revealed how different the panel’s approaches are to portfolio construction — and why both can work.
Harry revealed his team’s meeting load: “With my four investing partners, we have 80 net new companies that we meet in person per week.”
Jason’s response: “I would resign. I would give you all my carry back. I sold my companies because I didn’t want to spend my life in meetings. I’ll do one a week, two max.”
But Jason’s critique cuts deeper than meeting volume: “When you do 200 deals between your team, you got to meet 500, 600, 700 founders a year. I don’t want to meet 500 founders a year. Unless you’re great, you’re going to blow my mind. But if you’re not, I’m gonna start yawning about 15 minutes into this meeting.”
Rory defends the high-volume approach: “Even on an okay deal, you learn nuances about a specific market from the one-on-one that you wouldn’t get from the presentation. The person inside living it every day has a crucial kernel of knowledge that you just can’t access any other way.”
But the conversation reveals a deeper truth about fund construction:
The Small Fund, High Multiple Approach (Hummingbird-style):
- Keep fund size at $40-100M
- Accept dilution in follow-on rounds
- Start with 20% ownership, end with 12%
- Generate 8-10x returns
The Large Fund, Maintained Ownership Approach (Lightspeed-style):
- Raise $250M+ funds
- Deploy significant follow-on capital to maintain ownership
- Generate 5x returns but on much larger capital base
Rory’s key insight: “Both are great outcomes. They’re just different ways to play the game. The interesting thing is both outcomes have great outcomes for the GP. But if you only have $1 to play with as an LP, you want the small fund 10x. If you have to deploy $100M, you have to do the big fund. The high small fund accept the follow-on dilution but make a marvelous return is the compelling product for the marginal dollar.”
The Hummingbird Outlier: How to Actually Generate 10x Returns
Speaking of Hummingbird, their Billion to One investment deserves its own section because it represents what’s possible when everything goes right.
Hummingbird did their first bio deal in Billion to One and now has an $800M position on the IPO — likely representing an 800x+ return on their initial investment from a sub-$100M fund.
Harry marvels: “You want freaking great venture returns? Credit due. Amazing.”
But Jason asks the tactical question every seed investor should be asking: “How did they collect the capital as a seed manager to deploy enough to maintain the ownership? I don’t see how I’ll ever own 18% of something at IPO ever again.”
The answers are telling:
- Capital efficiency — Billion to One didn’t require massive rounds
- Follow-on investment — They did put in subsequent checks
- Concentration — They bet big on the winners
- Accepting some dilution — Rory suggests they may have gone from 20% to 12% but on a massive outcome
Rory makes the crucial point: “You can accept some dilution and optimize for multiple rather than ownership. If you put in $4M and get diluted from 20% to 12% but it’s a $5B exit, you’re a hero.”
Jason counters with what’s exciting about capital efficient outcomes: “Navan at $4.5B, Billion to One at $5B — the power of capital efficiency and running lean delivers superior investor returns despite smaller headline valuations.”
The lesson: Small funds generating 10x+ returns by accepting dilution but betting on capital-efficient companies may be the optimal strategy for seed investors who don’t have the deployment capacity of mega-funds.
What This All Means: The Real State of Venture in 2025
Pulling all these threads together reveals several uncomfortable truths:
1. The Middle Class of Fundable Companies is Gone
“It’s the most binary fundraising environment in our lifetimes,” Jason states flatly. The classic SaaS triple-triple-double-double company? A company growing from $400K to $3M ARR would have had five term sheets from good firms. Now: 120 meetings, one term sheet.
2. Most People from the Last Decade Should Step Aside
This isn’t ageism — it’s pace of change. “The most folks from the last decade or 15 years are not the right people for the next decade,” Jason argues. “I struggle to even recommend a lot of the CROs and executives I know for roles today. We don’t even need half the VCs we have today for the AI world.”
3. Diversification Matters More, Not Less
With increased variance in outcomes and compressed time to clone, the math says diversification is critical. But diversification at seed prices requires massive funds or acceptance of lower ownership.
4. There Are Only Three Ways to Win
Attach to compute budgets, replace human headcount, or massively displace incumbents. If you’re not doing one of those three things, you’re probably building a slow-growth SaaS company in a world that doesn’t want those anymore.
5. Capital Efficiency Delivers Superior Returns
Billion to One at $5B and Navan at $4.5B generated better investor returns than most $20B+ outcomes because they raised less capital. The lesson: optimize for investor returns, not headline valuations.
6. The Best Fundraises Don’t Feel Like Processes
Cultivate relationships over months. Have 3-4 investors ready to commit before you formally raise. Send one email and get multiple term sheets. That’s the optimal path.
7. If You’re Not Excited, You Should Retire
Jason’s most important point: “If this isn’t the most exciting time of your lifetime, you’re doing it wrong. This is the first time software has gotten better since the three of us met. If you’re not incredibly excited, retire. No shame in that. Put the rest into NASDAQ — you’re going to make more than most VC funds anyway.”
Looking Forward: What 2026 Brings
The panel is clear about what’s coming: “This year was AI works. Next year is AI is part of your team.”
That transition — from AI as a tool making you more efficient to AI as an actual member of your team with autonomy, knowledge, and capability — is where the revenue explosion happens.
Jason’s final thought captures the optimism underlying all the tough talk: “If you’re not incredibly excited right now, if this isn’t one of the most exciting and stressful times simultaneously, you’re missing it. This is a great enormous mega trend. It’s the biggest mega trend maybe since the early days of the internet. Leaning into it is the only sensible thing to do, and playing against it is dumb as rocks.”
The old playbook is broken. The new playbook is being written in real-time. And the investors who figure it out — whether through small concentrated funds like Hummingbird or large diversified portfolios like Harry’s approach — will capture the extraordinary returns available in this moment.
Everyone else? Maybe it’s time to buy that beach house.
Quotable Moments
On Venture Evolution:
“The most folks from the last decade or 15 years are not the right people for the next decade. I genuinely believe it across my ecosystem. I struggle to even recommend a lot of the CROs and executives I know for roles today.” — Jason Lemkin
“The pace of evolution is so fast. If you decide what I knew 6 months ago is still useful, you’re probably going to be wrong very quickly. That’s what I find the most stressful about right now.” — Rory O’Driscoll
On AI Investment Strategy:
“Tools are great when the AI is part of your team for real, not VC talk. The amount of revenue that it’s accessible is so high.” — Jason Lemkin
“Sell to the people who are making AI and if they grow, you’ll sell more too.” — Jason Lemkin
On Defensibility:
“You just can’t take that early first month explosion as seriously as you used to. It’s not as defensible.” — Harry Stebbings
“I don’t think you can have a major defensibility at the seed or even frankly the stage we’re investing at. The defensibility theorem emerges at scale.” — Rory O’Driscoll
On Market Timing:
“It’s easy to be roughly right. It’s very hard to imagine that the AI capex boom doesn’t have a significant correction. But going from that kind of armchair podcast statement to actually being able to make money on it — that’s damn hard.” — Rory O’Driscoll
On Fundraising:
“The best run processes don’t require a data room. Not a traditional one. Only one for diligence.” — Jason Lemkin
“The best run processes don’t feel like a process, but they are.” — Rory O’Driscoll
On Current Reality:
“It’s the most binary fundraising environment in our lifetimes. You’re either getting funded or you ain’t seeing it anywhere.” — Jason Lemkin
On Taking Action:
“If this isn’t the most exciting time of your lifetime, you’re doing it wrong. If you’re not incredibly excited, retire. No shame in that. You had a great run. Put the rest into NASDAQ — you’re going to make more than most VC funds anyway.” — Jason Lemkin
Listen to the full episode of 20VC x SaaStr on YouTube or your favorite podcast platform. For more insights on venture capital, AI investing, and B2B metrics, visit SaaStr.com and subscribe to the SaaStr newsletter.
